At 12:30 PM 3/23/2005 -0600, Boris Zbarsky wrote:
Peter Sorotokin wrote:
That would mean
- binding must happen before element is ever accessed (e.g at parse
time and in createElementNS or similar methods)
- bindings cannot be changed through styling
That seems to go against the direction of XBL2. We cannot have both
well-defined methods and dynamic binding attachment
The thing is, in any sort of reasonable system you can't have objects
randomly changing which interfaces they implement... The issue of binding
attachment through styling has been a major weak point of Mozilla XBL all
along, and causes so many problems that we are strongly considering moving
to a different, not style-related, binding mechanism altogether.
That's a good data point. This consideration was a primary motivation for
RCC to "bind" solely on the basis of the element name and namespace (which
sXBL retained). I was told that these kind of issues are never really a
problem in practice, but what you are saying indicates otherwise. However,
sXBL is a common activity with CSS group - and I wonder what that the
styling part of the TF would have to say. I agree that we need to have a
clear view on where we are headed in sXBL timeframe.
So while attaching bindings that don't define methods or implement
interfaces via style is a somewhat reasonable approach, bindings that
actually implement interfaces probably need to be attached via some other
mechanism.
That makes sense to me.
Peter
-Boris
